MDATA performance

Question

Hi,
I am contemplating the choice between MDATA and Field sections from performance point of view. My question is, if i have N (say 10) attributes of text, that change quite frequently, then which one of the two options will have better performance:
1. Changing (removing and adding) N MDATA section values.
2. Indexing the whole document where attributes are stored as Field sections.
To me, it appears that MDATA should give better performance, but how does performance depends on N ?
Thx,
RK

Answer 1

With the field section approach you are going to cause the index to get more fragmented (since you will be Sync'ing more often) and therefore affect query performance.
I expect the update performance for MDATA to be at least as good (if not better) as the update performance for field sections because even if all you have in your document is the field sections you are updating, the same values need to be populated in the $I table as with MDATA approach. Make sure you use that CTX_DDL.ADD_MDATA overloaded procedure which allows you to update MDATA values in BATCH - this will make the PL/SQL overhead small as N increases.
Faisal